Temporal and Spatial Patterns in Distemper Virus Cases Reported to SCWDS, 1975-2013
Jonathan Wilson
2019-12-11
Summary
Canine distemper is an important infectious disease that affects many mammal species. There is evidence of CDV infection in all terrestrial carnivores families and some marine carnivore families. CDV has been detected in wild animal species such as the African lion and Amur tigers and has been responsible for substantial population declines of the animals during outbreaks. Analysis was conducted on passive surveillance necropsy diagnosed wild animal cases presented to the South-Eastern Cooperative Wildlife Disease Study between 1975 and 2013. These analyses demonstrated the ability to use an ARIMA model to predict the numbers of Raccoon cases in a given month using data on the number of Gray Foxes cases from the previous three months. There also appears to be temporal trends in cases in both species, with cases more likely to occur during the breeding season. Spatially there tends to be clustering of cases of both species within the same areas, with cases appearing more likely to occur in area of medium human population density with fewer cases in very densely and sparsely populated areas, which may correspond to suburban areas where there is likely more contact with domestic dogs.
Introduction
Canine distemper is an important infectious disease that affects many mammal species. It is recognized as having the second highest case fatality rate of canine diseases , behind rabies (Swango, 1995). The causative agent, canine distemper virus (CDV) is an enveloped, single stranded, negative sense RNA virus in the Morbillivirus family. The major route of transmission is through aerosolization of virus containing respiratory exudate (Deem, Sharon L.; Spelman, L. H.; Yates, R. A.; Montali, 2000). CDV is highly infectious and virus may be shed for 60-90 days post infection (Greene, C. E. & Appel, 1990).
Distemper pathogenesis is well characterized in domestic dogs, with infections in wild species following a similar pattern. There is often a systemic infection with viremia and variable central nervous system involvement, dependent on the host’s immune response. Animals with a strong immune response will clear the virus and remain asymptomatic. A weak response will result in systemic disease and severe morbidity, often ultimately being fatal. An intermediate or delayed response results in delayed neurological signs and hyperkeratotic footpad lesions- the origin of the colloquial term “hard pad disease” (Greene, C. E. & Appel, 1990). The clinical signs of CDV infection are influenced by multiple factors, including; strain virulence, environmental conditions, host age and immune status and the animal species infected (Deem, Sharon L.; Spelman, L. H.; Yates, R. A.; Montali, 2000).
CDV results in a highly immunizing, acute infection. This type of pathogen normally requires high densities and large populations of hosts for long‐term persistence. CDV has been shown to be able to survive in wild carnivores due to multihost transmission, which overcomes the obstacle of population size and density within species (Almberg, Cross, & Smith, 2010). There is evidence of CDV infection in all terrestrial carnivores families and some marine carnivore families (Deem, Sharon L.; Spelman, L. H.; Yates, R. A.; Montali, 2000). Morbidity and mortality varies depending on the species but closely resembles rabies in wild carnivores (Hoff, Bigler, Proctor, & Stallings, 1974). The Mustelidae family is among the species with the highest fatality rate, while the domestic dog can be a asymptomatic carrier (Deem, Sharon L.; Spelman, L. H.; Yates, R. A.; Montali, 2000). CDV has been detected in wild animal species such as the African lion and Amur tigers and has been responsible for substantial population declines of the animals during outbreaks (Roelke-Parker et al., 1996; Seimon et al., 2013).
CDV is likely to have broad consequences for the health of both free-living and domestic carnivore species (R. J. Montali, G. R. Bartz &., 1987). CDV’s similarity to the measles virus suggest a shared ancestor and a zoonotic potential for CDV.
Canine Distemper Virus in the U.S.
CDV is seen most commonly in domestic cats and dogs, but frequent cross species transmission does occur in non-domestic carnivores (Greene, C. E. & Appel, 1990). Severity in domestic dogs depends on the animals’ immune status and age in addition to strain virulence (Beineke, Baumgärtner, & Wohlsein, 2015). In the US, raccoons (Procyon lotor), foxes (Vulpes vulpes and Urocyon cinereoargenteus), coyotes (Canis latrans), wolves (Canis lupus) , skunks (Mephitis mephitis), badgers (Taxidea taxus), mink (Mustela vison) and ferrets (Mustelidae spp.) are among the species susceptible to CDV infection(Kapil et al., 2008). In domestic ferrets mortality rates can reach 100% (Deem, Sharon L.; Spelman, L. H.; Yates, R. A.; Montali, 2000). CDV has been responsible for population declines of endangered mustelids like the black-footed ferret (Williams, Thorne, Appel, & Belitsky, 1988). CDV is also endemic in the eastern U.S. raccoon population. Raccoons are thought to be a reservoir for other wild animals and domestic dogs as well as other species of carnivores (Roscoe, 1993). CDV has been found to persist in areas like Yellow Stone national park, which has a diverse carnivore population. Multiple outbreaks have occurred in the wolf, coyote and cougar populations (Almberg et al., 2010; Almberg, Mech, Smith, Sheldon, & Crabtree, 2009).
It has been shown that Raccoons are able to maintain higher density populations within urban and suburban environments (Prange, Gehrt, & Wiggers, 2003). It was also demonstrated that these higher density populations are more likely to suffer mortality due to infectious disease. As distemper is a density dependent disease, it is likely that this disease incidence is influenced by urban vs rural environments.
Although raccoons are thought to be a major reservoir for CDV, little research has been done to identify the disease dynamics within this population. Available data is sparse, dated and focuses on individual states and discrete sites. CDV has also been shown to be a major cause of disease in the south eastern USS, with 78% of necropsy cases being diagnosed with the disease in one study (Davidson, Nettles, Hayes, Howerth, & Couvillion, 1992). The objective of this study is to identify spatial and temporal patterns in CDV cases, in a range of wild mammals, reported to the Southeastern Cooperative Wildlife Disease Study from 1975 to 2013.
Description of data and data source
Data was recorded from Canine distemper positive cases submitted to the Southeastern Cooperative Wildlife Disease Study (SCWDS) between 1975 and 2013. Cases were identified as CDV by fluorescent antibody testing and/or histologic diagnosis of characteristic lesions. Species, date of submission, county of origin, and sex were noted. A total number of 701 positive cases were submitted from 13 states over the 38-year period. Positive cases were comprised of raccoons (n=462), gray foxes (n=211), striped skunks (n=13), coyotes (n=7), red foxes (n=3), gray wolves (n=3), one mink and a black bear.
Census and county land area data for Georgia was accessed and downloaded from census.gov.
Questions to be addressed
1.Are there temporal trends in cases diagnosed? These may be longer term trends or seasonal trends related to breeding seasons. I hypothesize that there will be increased likelihood of cases occurring during the breeding season of wild mammals compared to the rest of the year.
2.Are there patterns in the timing of species being diagnosed suggesting cross species infection? Raccoons are considered primary reservoirs, are peaks in raccoon cases followed by cases in other species suggesting spillover? I hypothesize that there will be peaks in numbers of Raccoon CDV cases that are followed by cases in other species.
3.Are there spatial patterns of infection within the southeast? These patterns may be related to human activities and population density. Prevalence of CDV is reported to be higher in suburban foxes (Frölich et al., 2000) I expect that there will be a greater likelihood of cases in areas of higher human population density compared to lower density.
Methods
Data acquisition
Data of animals brought to SCWDS between 1975 and 2013, which were diagnosed as having CDV at post mortem. Cases were identified as CDV by fluorescent antibody testing and/or histologic diagnosis of characteristic lesions. Species, date of submission, county of origin, and sex were noted. This data was provided by SCWDS. Census and county land area data for Georgia was accessed and downloaded from census.gov.
Overview of Data
CDV case data contained the following variables; Case number, State, county, area, Sex, Species, Age and collection year. Additional data including specific collection dates was also used from a separate spread sheet for the time series analysis. The census data contained human population, land area and population density data for each county in the state of Georgia.
Data import and cleaning
Data was imported into R Studio and cleaned to correct data entry errors and missing data. Detailed description of data analysis and cleaning is contained in the processing scripts within the project repository in the processing_code subfolder.
Exploratory Analysis
All analyses were conducted in Program R version 3.5.3. The extent of exploratory analysis performed is outlined in the “ExploratoryAnalysis” file within the project repository.
Mapping
Mapping data for US states and Georgia counties is available through the ggplot2 (Wickham, 2016) package. As the CDV data only contained county level location information, county centroid coordinates were used for plotting case points.
Analysis of spatial spread of cases in Georgia was done using the spatstat package.(A. Baddeley & R.Turner., 2015) Analysis of the level of clustering of cases was done using Ripley’s K analysis. Ripley’s K looks at the level of randomness of spatial points. It analyses whether a point is more or less likely to be near another point than to a randomly plotted point.
Time Series Analysis
Time series analysis and ARIMA model construction was conducted using the “fpp2” package from Forecasting: principles and practice, Hyndman & Athanasopoulos.(Hyndman, R.J., & Athanasopoulos, 2018)
The three components of an ARIMA model are; auto-regression, differencing and the moving average. These translate as; autoregression uses previous data in the time series from n number of time points previously to predict future data, differencing computes the difference between observations in non-stationary data to remove the influence of trends or seasonality and the moving average uses the past forecast errors in the model to make future predictions. In addition to these basic components of the ARIMA model, lagged predictors were also included in model building. In the case of the Raccoon predictive model the lagged predictors were current and past Gray Fox cases (up to three months previously) and vice versa for the Gray Fox model.
For time series analysis the data was pooled into Raccoon or Gray Fox cases per month from April 1975-September 2013, for the whole state of Georgia. For training and testing of the ARIMA models for predicting Raccoon and Gray Fox cases per month, the sequence of months from April 1975-September 2013 (n=462) was divided at into the training set, up to n=368 ie. November 2005, and the testing set from n=369 to n=462, ie. December 2005-September 2013. The stationarity of the time series were confirmed using both the augmented Dickey–Fuller test (ADF-test) and the Kwiatkowski–Phillips–Schmidt–Shin test (KPSS-test), suggesting no overall trends or seasonality in the data. AutoARIMA analysis was performed on the time series using the forecast package.
Results
Data Exploration
Figure 1:Number of cases of CDV per state, submitted to SCWDS, 1975-2013.
A total of 16 states clustered in the south east were represented in the data set with Georgia containing the majority of cases (n=422).
Figure 2:CDV cases per state submitted to SCWDS, 1975-2013 .
The states represented were clustered in the south east of the US, with Pennsylvania the most northerly and Kansas the most westerly states.
Figure 3:Number of CDV cases per species submitted to SCWDS, 1975-2013
There are eight species represented, with Gray Foxes (n=211) and Raccoons (n=462) making up the majority of cases.
Figure 4:Total number of cases of CDV diagnosed at SCWDS per Year, 1975-2013.
Every year between 1975 and 2013 had at least one diagnosed case of CDV.
Initial probing of the data set revealed the vast majority of cases to be submitted from the state of Georgia. The other feature is that almost all of the submitted cases are Raccoons or Gray Foxes. From this point, data exploration and analysis will focus only on Gray foxes and Raccoons in the state of Georgia as this compromises the majority of cases.
Figure 5: Species and Age of CDV cases submitted to SCWDS, 1975-2013.
Despite the Chi-squared value (7.686, p=0.02143) being greater than the critical value in this case, the age data is difficult to use as it is a particularly subjective measure in this case and there would need to be very marked changes for any relationships to be suggested.
Figure 6: Species and Sex of CDV cases submitted to SCWDS, 1975-2013.
The Chi-squared value of Sex and Species did not reach the critical so the null hypothesis that sex and species are independent can be accepted in this case.
Figure 7: Raccoon and Gray Fox cases in Georgia presented to SCWDS, 1975-2013.
Raccoon and Gray Foxes cases between 1975 and 2013 showed significant fluctuation with the 1980s having the greatest case numbers per year for both species. There were only four years in which the number of diagnosed Gray Fox cases exceeded the number of Raccoon cases.
Figure 8: Raccoon and Gray Fox cases across breeding cycle.
In this case the different parts of the reproduction cycle were defined as the Breeding season being February to April, Lactation as May to July and the rest of the year non-Breeding. The number of cases recorded in each of these parts of the reproductive cycle were adjusted by dividing by the length of the segment in months to account for any differences in the number of cases due to the length of the recording period.
Spatial Mapping of Georgia Data
Further qualitative analysis was conducted by mapping disease presence over time at county level.
Figure 9:Total Raccoon and Gray Fox cases in Georgia presented to SCWDS, 1975-2013, per county.
Cases appear to be clustered in the northern part of the state with a second smaller cluster in the south east.
Figure 10:Raccoon and Gray Fox cases in Georgia in the Northern and Southern halves of the state, presented to SCWDS, 1975-2013.
The cases were divided into North and South by using the central point of the state, which is in Twiggs County, 32.67328 latitude. Chi-squared analysis was performed using the null hypothesis that species and northern and southern latitudes are independent. The Chi-squared value was 6.9591, which is above the critical value with a p-value of 0.008339 so the null hypothesis can be rejected. In this data set it appears that Gray Fox cases are more likely to occur in the northern part of the state